In general, speech recognition systems can use a neural network model that performs speech enhancement and acoustic modeling. Some systems process audio data from multiple microphones using a neural network.